Pub Date : 2025-12-26DOI: 10.1016/j.isci.2025.114546
Di Wang , Mingyu Yang , Yuhan Liu , Xu Hu , Feiyu Hao , Daqian Sun , Nan Wang , Yuanyuan Dong , Weican Liu , Xiaowei Li , Fawei Wang
Soybean (Glycine max (L.) Merr.) is an economically important crop. Drought adversely affects nutrient acquisition and early vegetative development, thereby constraining the growth potential of soybean seedlings. Phospholipase C (PLC) participates in diverse biological processes, yet its function in soybeans has not been fully explored. In this study, we generated GmPLC8-overexpressing (OE) and GmPLC8-silenced soybean seedlings to investigate the function of GmPLC8. The results showed that GmPLC8-OE lines exhibited higher tolerance than wild-type seedlings, whereas GmPLC8-silenced seedlings were more sensitive. This was associated with reduced malondialdehyde (MDA) and hydrogen peroxide (H2O2) levels in GmPLC8-OE lines, and increased levels in GmPLC8-silenced seedlings. Further analysis revealed that drought caused smaller stomatal apertures and higher abscisic acid (ABA) content in GmPLC8-OE lines. Moreover, the transcription levels of ABA-related genes were upregulated in GmPLC8-OE lines. Taken together, these findings suggest that GmPLC8 enhances drought tolerance in soybean by regulating stomatal aperture.
{"title":"GmPLC8 enhances drought resistance in soybean by regulating stomatal closure","authors":"Di Wang , Mingyu Yang , Yuhan Liu , Xu Hu , Feiyu Hao , Daqian Sun , Nan Wang , Yuanyuan Dong , Weican Liu , Xiaowei Li , Fawei Wang","doi":"10.1016/j.isci.2025.114546","DOIUrl":"10.1016/j.isci.2025.114546","url":null,"abstract":"<div><div>Soybean (<em>Glycine max</em> (L.) Merr.) is an economically important crop. Drought adversely affects nutrient acquisition and early vegetative development, thereby constraining the growth potential of soybean seedlings. Phospholipase C (PLC) participates in diverse biological processes, yet its function in soybeans has not been fully explored. In this study, we generated <em>GmPLC8</em>-overexpressing (OE) and <em>GmPLC8</em>-silenced soybean seedlings to investigate the function of <em>GmPLC8</em>. The results showed that <em>GmPLC8</em>-OE lines exhibited higher tolerance than wild-type seedlings, whereas <em>GmPLC8</em>-silenced seedlings were more sensitive. This was associated with reduced malondialdehyde (MDA) and hydrogen peroxide (H<sub>2</sub>O<sub>2</sub>) levels in <em>GmPLC8</em>-OE lines, and increased levels in <em>GmPLC8</em>-silenced seedlings. Further analysis revealed that drought caused smaller stomatal apertures and higher abscisic acid (ABA) content in <em>GmPLC8</em>-OE lines. Moreover, the transcription levels of ABA-related genes were upregulated in <em>GmPLC8</em>-OE lines. Taken together, these findings suggest that <em>GmPLC8</em> enhances drought tolerance in soybean by regulating stomatal aperture.</div></div>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 2","pages":"Article 114546"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146035713","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26DOI: 10.1016/j.isci.2025.114407
Yongze Song , Filip Biljecki , Gustau Camps-Valls , Peter M. Atkinson
Geospatial artificial intelligence (GeoAI) is reshaping our understanding of Earth and urban systems by integrating advanced artificial intelligence techniques with diverse geospatial data and methodologies. This backstory highlights recent GeoAI advances and applications as presented in the 11 articles in the iScience special issue, “GeoAI shaping earth and cities: Advances, opportunities, and challenges.” Guest editors share perspectives on GeoAI’s advances in explainability, adaptability, and sustainability, demonstrating that GeoAI’s applications extend beyond traditional mapping functions. These 11 case studies illustrate four types of explainability, three levels of adaptability and three thematic areas of sustainability, showing the methodological diversity and practical relevance of GeoAI for Earth and urban systems. Here, interactions among these dimensions are mapped to support the evaluation and design of future GeoAI solutions. We also outline future research directions for GeoAI to address complex challenges across the sciences relating to the Earth and its cities.
{"title":"GeoAI: Beyond mapping earth and cities through explainability, adaptability, and sustainability","authors":"Yongze Song , Filip Biljecki , Gustau Camps-Valls , Peter M. Atkinson","doi":"10.1016/j.isci.2025.114407","DOIUrl":"10.1016/j.isci.2025.114407","url":null,"abstract":"<div><div>Geospatial artificial intelligence (GeoAI) is reshaping our understanding of Earth and urban systems by integrating advanced artificial intelligence techniques with diverse geospatial data and methodologies. This backstory highlights recent GeoAI advances and applications as presented in the 11 articles in the <em>iScience</em> special issue, “GeoAI shaping earth and cities: Advances, opportunities, and challenges.” Guest editors share perspectives on GeoAI’s advances in explainability, adaptability, and sustainability, demonstrating that GeoAI’s applications extend beyond traditional mapping functions. These 11 case studies illustrate four types of explainability, three levels of adaptability and three thematic areas of sustainability, showing the methodological diversity and practical relevance of GeoAI for Earth and urban systems. Here, interactions among these dimensions are mapped to support the evaluation and design of future GeoAI solutions. We also outline future research directions for GeoAI to address complex challenges across the sciences relating to the Earth and its cities.</div></div>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 1","pages":"Article 114407"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145837615","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26DOI: 10.1016/j.isci.2025.114549
Martin Lang , Alexandra Ružičková , Katarína Čellárová , Eva Kundtová Klocová , Radim Chvaja
While costly signals of commitment were proposed to facilitate cooperative assortment in humans, little is known about cognitive processes facilitating the reliability of such commitment signals. Across three studies (total N = 2,223), we test whether this reliability is supported by intuitive processes or by deliberative cost-benefit evaluations. Participants played a public goods game in which they could send variously costly signals of commitment, and the signaling decision was made either under time pressure or delay. Under pressure, signals separated cooperators from selfish players only when signal cost was high. However, these intuitive, high-cost signals did not consistently predict subsequent cooperation. By contrast, when participants had time to deliberate under time delay, cooperators selectively chose the low-cost signal more often than selfish players and contributed more to the common pool. These findings suggest that reliable commitment signaling depends less on intuitive attraction to costly displays and more on deliberative cost-benefit evaluation where cooperators optimize signal cost.
{"title":"Intuitive and deliberative processes underlying commitment signaling for cooperative assortment","authors":"Martin Lang , Alexandra Ružičková , Katarína Čellárová , Eva Kundtová Klocová , Radim Chvaja","doi":"10.1016/j.isci.2025.114549","DOIUrl":"10.1016/j.isci.2025.114549","url":null,"abstract":"<div><div>While costly signals of commitment were proposed to facilitate cooperative assortment in humans, little is known about cognitive processes facilitating the reliability of such commitment signals. Across three studies (total <em>N</em> = 2,223), we test whether this reliability is supported by intuitive processes or by deliberative cost-benefit evaluations. Participants played a public goods game in which they could send variously costly signals of commitment, and the signaling decision was made either under time pressure or delay. Under pressure, signals separated cooperators from selfish players only when signal cost was high. However, these intuitive, high-cost signals did not consistently predict subsequent cooperation. By contrast, when participants had time to deliberate under time delay, cooperators selectively chose the low-cost signal more often than selfish players and contributed more to the common pool. These findings suggest that reliable commitment signaling depends less on intuitive attraction to costly displays and more on deliberative cost-benefit evaluation where cooperators optimize signal cost.</div></div>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 2","pages":"Article 114549"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26DOI: 10.1016/j.isci.2025.114553
Sam LaMagna , Eduardo Solessio
Gap junctions are important components of signal communication in sensory systems. In the retina, connexin 36 (Cx36) gap junctions integrate visual signals over space and time. However, we do not understand how Cx36 gap junctions in the inner- and outer-retina govern visual thresholds under dim light conditions. Here we address this question using transgenic mouse models, an operant behavioral assay, and psychophysical modeling. Visual thresholds with Ganzfeld illumination decreased with flash durations <0.5 s, plateauing at ∼2 × 10−4 ph/μm2/s with longer flashes. Removal of Cx36 in the outer retina lowered visual thresholds by 2.6-fold. However, removing Cx36 in both the inner and outer retina increased visual thresholds by 16.5-fold relative to controls. Further, we demonstrated that the visual thresholds of Cx36KO mice are rod-mediated, most likely via the OFF-rod pathway. We conclude that both inner- and outer-retinal Cx36 control visual thresholds under dim light conditions, but in distinct ways.
{"title":"Retinal Cx36 gap junctions in the inner and outer plexiform layers differentially control visual thresholds","authors":"Sam LaMagna , Eduardo Solessio","doi":"10.1016/j.isci.2025.114553","DOIUrl":"10.1016/j.isci.2025.114553","url":null,"abstract":"<div><div>Gap junctions are important components of signal communication in sensory systems. In the retina, connexin 36 (Cx36) gap junctions integrate visual signals over space and time. However, we do not understand how Cx36 gap junctions in the inner- and outer-retina govern visual thresholds under dim light conditions. Here we address this question using transgenic mouse models, an operant behavioral assay, and psychophysical modeling. Visual thresholds with Ganzfeld illumination decreased with flash durations <0.5 s, plateauing at ∼2 × 10<sup>−4</sup> ph/μm<sup>2</sup>/s with longer flashes. Removal of Cx36 in the outer retina lowered visual thresholds by 2.6-fold. However, removing Cx36 in both the inner and outer retina increased visual thresholds by 16.5-fold relative to controls. Further, we demonstrated that the visual thresholds of Cx36KO mice are rod-mediated, most likely via the OFF-rod pathway. We conclude that both inner- and outer-retinal Cx36 control visual thresholds under dim light conditions, but in distinct ways.</div></div>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 2","pages":"Article 114553"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146035795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26eCollection Date: 2026-01-16DOI: 10.1016/j.isci.2025.114545
Zenebe Markos Lonseko, Helen Haile Hayeso, Tan Gan, Jinlin Yang, Nini Rao
Accurate and early diagnosis of esophageal cancer (EC) remains challenging due to heterogeneous imaging characteristics and the complexity of integrating multimodal data. This study proposes a multimodal-multiscale hybrid fusion network (MHF-Net), which combines domain-driven handcrafted features (HFs) and deep multiscale features (DFs) from multimodal data for the automated diagnosis of EC. MHF-Net uses a dilated-inception block to extract multiscale representations at varied dilation rates and concatenates these DF with domain-specific features via dense connections. A convolutional block attention module further refined spatial and channel-wise features, leveraging global average pooling. The grasshopper optimization algorithm optimized fusion weights and hyperparameters, enhancing hybrid feature integration and overall model robustness. Evaluated across five diverse datasets, MHF-Net achieves state-of-the-art performance (accuracy = 0.970 ± 0.014; F1 score = 0.960 ± 0.017). This study demonstrates clinical applicability with strong potential for future enhancement through the integration of multimodal biomarkers for enhanced diagnostic accuracy.
{"title":"Multimodal-multiscale hybrid fusion network for automated esophageal cancer diagnosis.","authors":"Zenebe Markos Lonseko, Helen Haile Hayeso, Tan Gan, Jinlin Yang, Nini Rao","doi":"10.1016/j.isci.2025.114545","DOIUrl":"10.1016/j.isci.2025.114545","url":null,"abstract":"<p><p>Accurate and early diagnosis of esophageal cancer (EC) remains challenging due to heterogeneous imaging characteristics and the complexity of integrating multimodal data. This study proposes a multimodal-multiscale hybrid fusion network (MHF-Net), which combines domain-driven handcrafted features (HFs) and deep multiscale features (DFs) from multimodal data for the automated diagnosis of EC. MHF-Net uses a dilated-inception block to extract multiscale representations at varied dilation rates and concatenates these DF with domain-specific features via dense connections. A convolutional block attention module further refined spatial and channel-wise features, leveraging global average pooling. The grasshopper optimization algorithm optimized fusion weights and hyperparameters, enhancing hybrid feature integration and overall model robustness. Evaluated across five diverse datasets, MHF-Net achieves state-of-the-art performance (accuracy = 0.970 ± 0.014; F1 score = 0.960 ± 0.017). This study demonstrates clinical applicability with strong potential for future enhancement through the integration of multimodal biomarkers for enhanced diagnostic accuracy.</p>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 1","pages":"114545"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12818140/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146016806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26eCollection Date: 2026-01-16DOI: 10.1016/j.isci.2025.114543
Uma M Lal-Trehan Estrada, Arnau Oliver, Sunil A Sheth, Xavier Lladó, Luca Giancardo
[This corrects the article DOI: 10.1016/j.isci.2024.108881.].
[这更正了文章DOI: 10.1016/j.i ssn .2024.108881.]。
{"title":"Erratum: Strategies to combine 3D vasculature and brain CTA with deep neural networks: Application to LVO.","authors":"Uma M Lal-Trehan Estrada, Arnau Oliver, Sunil A Sheth, Xavier Lladó, Luca Giancardo","doi":"10.1016/j.isci.2025.114543","DOIUrl":"https://doi.org/10.1016/j.isci.2025.114543","url":null,"abstract":"<p><p>[This corrects the article DOI: 10.1016/j.isci.2024.108881.].</p>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 1","pages":"114543"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12797028/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145970322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26eCollection Date: 2026-01-16DOI: 10.1016/j.isci.2025.114556
Yuichiro Adachi, Allison M Williams, Masashi Masuda, Yutaka Taketani, Paul J Anderson, Pavel Ivanov
Stress granules (SGs) are cytoplasmic ribonucleoprotein condensates formed in response to stress-induced inhibition of mRNA translation and polysome disassembly. Despite the broad interest in SG assembly and disassembly in response to acute stress, SG dynamics under chronic stress has not been extensively investigated. We show that cells pre-conditioned with low-dose chronic (24 h exposure) stresses of various natures fail to assemble SGs in response to acute stress. While protein synthesis is drastically decreased by acute stress in pre-conditioned cells, polysome profiling analysis reveals the partial preservation of polysomes. Mechanistically, chronic stress slows down the rate of mRNA translation at the elongation phase, and triggers phosphorylation of translation elongation factor eEF2. These events further promote ribosome stalling, which is distinct from ribosome collisions known to trigger ribosome-associated quality-control pathways. In summary, chronic stress triggers ribosome stalling, which prevents efficient polysome disassembly and SG formation by subsequent acute stress.
{"title":"Chronic stress antagonizes formation of stress granules.","authors":"Yuichiro Adachi, Allison M Williams, Masashi Masuda, Yutaka Taketani, Paul J Anderson, Pavel Ivanov","doi":"10.1016/j.isci.2025.114556","DOIUrl":"10.1016/j.isci.2025.114556","url":null,"abstract":"<p><p>Stress granules (SGs) are cytoplasmic ribonucleoprotein condensates formed in response to stress-induced inhibition of mRNA translation and polysome disassembly. Despite the broad interest in SG assembly and disassembly in response to acute stress, SG dynamics under chronic stress has not been extensively investigated. We show that cells pre-conditioned with low-dose chronic (24 h exposure) stresses of various natures fail to assemble SGs in response to acute stress. While protein synthesis is drastically decreased by acute stress in pre-conditioned cells, polysome profiling analysis reveals the partial preservation of polysomes. Mechanistically, chronic stress slows down the rate of mRNA translation at the elongation phase, and triggers phosphorylation of translation elongation factor eEF2. These events further promote ribosome stalling, which is distinct from ribosome collisions known to trigger ribosome-associated quality-control pathways. In summary, chronic stress triggers ribosome stalling, which prevents efficient polysome disassembly and SG formation by subsequent acute stress.</p>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 1","pages":"114556"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12828549/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146049621","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26eCollection Date: 2026-01-16DOI: 10.1016/j.isci.2025.114548
Alejandro de Janon, Pranitha Kaza, Kaede Yoshida, SooJung Lee, Deborah DeRyckere, Douglas K Graham, Shuichi Takayama
The vasculature is essential for tissue function and pathology. Spheroid co-cultures of endothelial and marrow/mesenchymal stromal/stem cells (MSCs) form consistent structures, but the vascular components are short-lived. iPSC-derived vascular organoids can establish complex vasculature but often have variable cell maturation and low reproducibility. This article presents consistently formed, free-floating, long-term vascularized mesenchymal organoids (VMOs), by co-culturing human umbilical vein endothelial cells (HUVECs) and MSCs in a pre-gelled minimal Matrigel scaffold. VMOs support 60-day stable vasculature, exhibiting tissue maturation involving inflammation, extracellular matrix remodeling, and endothelial development. Compared to traditional spheroids, VMOs showed enhanced vascular complexity, sustained extracellular matrix production, and higher cell viability. The system preserved MSC heterogeneity including perivascular cell types, offering physiological relevance. Engraftment of breast cancer cells revealed stromal-tumor niches, enabling modeling of bone marrow metastasis. This robust platform offers an alternative model for studying vascular biology, stromal dynamics, and cancer progression, with potential applications in drug testing.
{"title":"Free-floating long-term vascularized mesenchymal organoids.","authors":"Alejandro de Janon, Pranitha Kaza, Kaede Yoshida, SooJung Lee, Deborah DeRyckere, Douglas K Graham, Shuichi Takayama","doi":"10.1016/j.isci.2025.114548","DOIUrl":"10.1016/j.isci.2025.114548","url":null,"abstract":"<p><p>The vasculature is essential for tissue function and pathology. Spheroid co-cultures of endothelial and marrow/mesenchymal stromal/stem cells (MSCs) form consistent structures, but the vascular components are short-lived. iPSC-derived vascular organoids can establish complex vasculature but often have variable cell maturation and low reproducibility. This article presents consistently formed, free-floating, long-term vascularized mesenchymal organoids (VMOs), by co-culturing human umbilical vein endothelial cells (HUVECs) and MSCs in a pre-gelled minimal Matrigel scaffold. VMOs support 60-day stable vasculature, exhibiting tissue maturation involving inflammation, extracellular matrix remodeling, and endothelial development. Compared to traditional spheroids, VMOs showed enhanced vascular complexity, sustained extracellular matrix production, and higher cell viability. The system preserved MSC heterogeneity including perivascular cell types, offering physiological relevance. Engraftment of breast cancer cells revealed stromal-tumor niches, enabling modeling of bone marrow metastasis. This robust platform offers an alternative model for studying vascular biology, stromal dynamics, and cancer progression, with potential applications in drug testing.</p>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 1","pages":"114548"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12818071/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146016803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26DOI: 10.1016/j.isci.2025.114555
Medard Edmund Mswahili , JunHa Hwang , Kyuri Jo , Jagath C. Rajapakse , Péter Kovács , Young-Seob Jeong
Efficient molecular representations are critical for improving the performance and generalization of large language models in chemical learning. Transformer-based architectures have advanced molecular representation learning, yet capturing localized and hierarchical chemical structures remains challenging. We introduce the SMILES Byte-Patch Transformer (SMiBPT), an adaptive model that dynamically segments SMILES and DeepSMILES strings into chemically meaningful substructures through entropy-based byte patching. Unlike fixed and traditional tokenization, our method adjusts patch sizes via entropy thresholds to preserve molecular semantics and enhance feature extraction. SMiBPT integrates self-supervised pretraining, chemical motif-aware encoding, adaptive entropy-aware masking, and rotary position embeddings to model ∼216 million unlabeled molecules from PubChem without truncation. We analyze entropy patterns across molecular classes and optimize patching by chemical properties such as aromaticity, charge, and functional groups. Through large-scale pretraining on ∼1.2 billion patches and extensive ablation studies, SMiBPT, despite its parameter-efficient transformer architecture, outperforms ChemBERTa, SMILES-BERT, and MoLFormer in predictive accuracy, efficiency, and zero-shot transfer analysis.
{"title":"Entropy-based byte patching transformer for self-supervised pretraining of SMILES data","authors":"Medard Edmund Mswahili , JunHa Hwang , Kyuri Jo , Jagath C. Rajapakse , Péter Kovács , Young-Seob Jeong","doi":"10.1016/j.isci.2025.114555","DOIUrl":"10.1016/j.isci.2025.114555","url":null,"abstract":"<div><div>Efficient molecular representations are critical for improving the performance and generalization of large language models in chemical learning. Transformer-based architectures have advanced molecular representation learning, yet capturing localized and hierarchical chemical structures remains challenging. We introduce the SMILES Byte-Patch Transformer (SMiBPT), an adaptive model that dynamically segments SMILES and DeepSMILES strings into chemically meaningful substructures through entropy-based byte patching. Unlike fixed and traditional tokenization, our method adjusts patch sizes via entropy thresholds to preserve molecular semantics and enhance feature extraction. SMiBPT integrates self-supervised pretraining, chemical motif-aware encoding, adaptive entropy-aware masking, and rotary position embeddings to model ∼216 million unlabeled molecules from PubChem without truncation. We analyze entropy patterns across molecular classes and optimize patching by chemical properties such as aromaticity, charge, and functional groups. Through large-scale pretraining on ∼1.2 billion patches and extensive ablation studies, SMiBPT, despite its parameter-efficient transformer architecture, outperforms ChemBERTa, SMILES-BERT, and MoLFormer in predictive accuracy, efficiency, and zero-shot transfer analysis.</div></div>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 2","pages":"Article 114555"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146035928","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-26DOI: 10.1016/j.isci.2025.114575
Kevin C. Yu , Alex I. Wiesman , Elizabeth M. Davenport , Laura A. Flashman , Jillian E. Urban , Srikantan S. Nagarajan , Kiran K. Solingpuram Sai , Joel D. Stitzel , Joseph A. Maldjian , Christopher T. Whitlow
American tackle football is associated with high rates of concussion, leading to neurophysiological disturbances and debilitating clinical symptoms, but previous investigations have largely ignored aperiodic neurophysiological activity. In this study, we use pre- and post-season resting-state magnetoencephalography (MEG) to examine whether concussion during a season of football is related to changes in neurophysiology and whether any such changes are associated with clinical outcomes. Concussion was associated with increased aperiodic exponents in superior frontal cortices, indicating a relative reduction in cortical excitability, which accounted for effects on raw delta and gamma power and was associated with worse cognitive concerns across participants. Subtler effects of non-concussive head-impact exposure on neurophysiology were similar to those of concussion. These findings indicate that concussion alters the excitability of the cortex, similar to much more subtle effects of non-concussive exposure.
{"title":"Reduced cortical excitability is associated with head impact severity and cognitive symptoms in adolescent football players","authors":"Kevin C. Yu , Alex I. Wiesman , Elizabeth M. Davenport , Laura A. Flashman , Jillian E. Urban , Srikantan S. Nagarajan , Kiran K. Solingpuram Sai , Joel D. Stitzel , Joseph A. Maldjian , Christopher T. Whitlow","doi":"10.1016/j.isci.2025.114575","DOIUrl":"10.1016/j.isci.2025.114575","url":null,"abstract":"<div><div>American tackle football is associated with high rates of concussion, leading to neurophysiological disturbances and debilitating clinical symptoms, but previous investigations have largely ignored aperiodic neurophysiological activity. In this study, we use pre- and post-season resting-state magnetoencephalography (MEG) to examine whether concussion during a season of football is related to changes in neurophysiology and whether any such changes are associated with clinical outcomes. Concussion was associated with increased aperiodic exponents in superior frontal cortices, indicating a relative reduction in cortical excitability, which accounted for effects on raw delta and gamma power and was associated with worse cognitive concerns across participants. Subtler effects of non-concussive head-impact exposure on neurophysiology were similar to those of concussion. These findings indicate that concussion alters the excitability of the cortex, similar to much more subtle effects of non-concussive exposure.</div></div>","PeriodicalId":342,"journal":{"name":"iScience","volume":"29 2","pages":"Article 114575"},"PeriodicalIF":4.1,"publicationDate":"2025-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145975106","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号